Multiple parallel packet routing lookup

BACKGROUND OF THE INVENTION

1. Field of the Invention

The invention relates to packet switching.

2. Related Art

In a packet-switched network, a “router” is a device which receives packets on one or more input interfaces and which outputs those packets on one of a plurality of output interfaces, so as to move those packets within the network from a source device to a destination device. Each packet includes header information which indicates the destination device (and other information), and the router includes routing information which associates an output interface with information about the destination device (possibly with other information). The router can also perform other operations on packets, such as rewriting the packets according to their routing protocol or to reencapsulate the packets from a first routing protocol to a second routing protocol. It is advantageous for routers to operate as quickly as possible, so that as many packets as possible can be switched in a unit time.

One operation performed by routers is routing lookup, that is, accessing routing information in response to the header information from the packet. For example, the router can determine an output interface on which to output the packet in response to a destination address specified by the packet header. In some routing protocols such as IP, an entire set of destination addresses can be associated with a single output interface, so that the operation of routing lookup can be responsive to routing information of differing lengths.

A first problem which has arisen in the art is that storing and retrieving routing information can be both complex and slow, due to the number of differing ways in which the destination address or other packet header information can be associated with that routing information. For example, methods by which the router might perform routing lookup for longer sets of routing lookup information can be inefficient for shorter sets of routing lookup information, and methods by which the router might perform routing lookup for shorter sets of routing lookup information can be inefficient for longer sets of routing lookup information.

A second problem which has arisen in the art is that header information associated with routing lookup has been seen to follow identifiable patterns, particularly for IP addresses used in the internet. Among those identifiable patterns are that 21-bit and 24-bit destination address headers are relatively common, while 32-bit destination addresses and 8-bit destination address headers are relatively rare. Thus, methods by which the router might perform routing lookup should be efficient for lookup of middle-length destination addresses without being inefficient for lookup of relatively longer or shorter destination addresses.

Some known routers, such as those described in U.S. application Ser. No. 08/655,429, “Network Flow Switching and Flow Data Export”, filed May 28, 1996, in the name of inventors Darren Kerr and Barry Bruins, and assigned to Cisco Systems, Inc., and U.S. application Ser. No. 08/771,438, having the same title, filed Dec. 20, 1996, in the name of the same inventors, assigned to the same assignee, can perform routing lookup for differing length destination addresses, by successively performing routing lookup for successive bytes of the destination address. Thus, each byte of the destination address provides further information from which specific information for routing the packet can be addressed. While this method achieves the goal of being relatively flexible with regard to the length of the destination address required for routing lookup, it can take many clock cycles to perform routing lookup, and is therefore not as relatively quick as desired. Moreover, while this method is relatively efficient for relatively shorter length destination addresses, it becomes increasingly inefficient as the lengths of destination addresses become relatively longer.

Accordingly, it would be desirable to provide a method and system for performing routing lookup, which is responsive to a plurality of different sets of routing lookup information. This advantage is achieved in an embodiment of the invention in which a plurality of sets of routing lookup information are queued for lookup in an external memory, particularly where a plurality of sets of routing lookup information are distinguished both by packet routing information and to the length of that header information.

SUMMARY OF THE INVENTION

The invention provides a method and system for routing information lookup for packets using a routing protocol such as IP. Routing information is determined responsive to the packet header, which in a preferred embodiment includes a destination address, a source address, and an input interface for the packet. Routing lookup is performed in response to at least one set of selected routing information, using a lookup table which includes tags both for the routing information and for a bitmask length (thus indicating the generality or scope of the routing information for the routing lookup).

In a preferred embodiment, the lookup table is structured so that addresses having the most common bitmask length are addressed first, but that more specific addresses are still considered when they are present. It has been discovered that most internet addresses can be found by reference to 24-bit or 21-bit IP addresses, after which 16-bit, 12-bit, and finally 32-bit IP addresses are considered. Lookup flags indicate when a relatively uncommon but more specific 32-bit IP address match is available. A memory controller pipelines the lookup requests to a hash table memory, flushes superfluous requests when a lookup result is found, and handles cases relating to 32-bit IP address matches.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 shows a block diagram of a system for multiple parallel packet routing lookup.

FIG. 2 shows a process flow diagram of a method of operating a system for multiple parallel packet routing lookup.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT

In the following description, a preferred embodiment of the invention is described with regard to preferred process steps and data structures. Those skilled in the art would recognize after perusal of this application that embodiments of the invention can be implemented using general purpose processors or special purpose processors or other circuits adapted to particular process steps and data structures described herein, and that implementation of the process steps and data structures described herein would not require undue experimentation or further invention.

Inventions described herein can be used in conjunction with inventions described in the following applications:

application Ser. No. 08/918,505, filed the same day, Express Mail Mailing No. EM 337222825 US, in the name of the same inventor, titled “Multiprotocol Packet Recognition and Switching”, and

application Ser. No. 08/918,506, filed the same day, Express Mail Mailing No. EM 571204544 US, in the name of the same inventor, titled “Enhanced Internet Packet Routing Lookup”.

Each of these applications is hereby incorporated by reference as if fully set forth herein.

System for Enhanced Internet Packet Address Lookup

FIG. 1 shows a block diagram of a system for multiple parallel packet routing lookup.

A system <b>100</b> includes an input port <b>101</b> for receiving and distributing a plurality of m<b>1</b> words <b>102</b> of routing lookup information; in a preferred embodiment the routing lookup information comprises the plurality of m multiplexer output words <b>232</b> and the system <b>100</b> comprises the lookup memory element <b>131</b>, as shown by figures in the co-pending application “Multiprotocol Packet Recognition and Switching”, application Ser. No. 08/918,505.

In a preferred embodiment, the plurality of m<b>1</b> words <b>102</b> includes m<b>1</b> different sets of routing lookup information which include m<b>1</b> different ways in which routing results for an associated packet can be determined. For a first example, where the associated packet is transmitted using an IP multicast protocol such as IGMP, the m<b>1</b> words <b>102</b> can include both (s, g) and (*, g) lookup information for that packet. For a second example, where the associated packet is transmitted using an IP unicast protocol such as IGRP, the m<b>1</b> words <b>102</b> can include routing lookup information for that packet responsive to both the packet address and a network bitmask length, as described in further detail in the co-pending application “Enhanced Internet Packet Routing Lookup”, application Ser. No. 08/918,506.

In a preferred embodiment, m<b>1</b> is about ten, so as to accommodate about ten different ways in which routing results for an associated packet can be determined. Of these, in a preferred embodiment, five are used for IP version 4, two for IP multicast (IGMP), one for TAG switching, one for IPX, and one as a spare for expansion.

The plurality of m<b>1</b> words <b>102</b> are coupled to a lookup holding table <b>110</b>. The lookup holding table <b>110</b> includes a plurality of k<b>1</b> columns <b>111</b> and a plurality of m<b>2</b> rows <b>112</b>. An entry <b>113</b> is defined at the unique intersection of each column <b>111</b> and row <b>112</b>.

Each of the m<b>1</b> words <b>102</b> is coupled to a corresponding set of the m<b>2</b> rows <b>112</b> of the lookup holding table <b>110</b>. The correspondence between the m<b>1</b> words <b>102</b> and the m<b>2</b> rows <b>112</b> is not one-to-one; there are a greater number of the m<b>1</b> words <b>102</b> than the m<b>2</b> rows <b>112</b>. Thus, a first one of the m<b>2</b> rows <b>112</b> is coupled to all of the m<b>1</b> words <b>102</b>, a second one of the m<b>2</b> rows <b>112</b> is coupled to some of the m<b>1</b> words <b>102</b>, and a last one of the m<b>2</b> rows <b>112</b> is coupled to at least one of the m<b>1</b> words <b>102</b>.

In a preferred embodiment, m<b>2</b> is about five, so as to accommodate about five simultaneous matches for a selected protocol. Since the m<b>1</b> words <b>102</b> include five sets of lookup information for IP version 4, five is the expected maximum number of simultaneous matches.

Each entry <b>113</b> includes one word <b>102</b> of routing lookup information and a valid bit <b>114</b> indicating whether the word <b>102</b> in the entry <b>113</b> is meaningful. In a preferred embodiment where those words <b>102</b> can be of variable length, each entry <b>113</b> is sufficiently large to hold a largest possible word <b>102</b> of routing lookup information, with unused parts of the entry <b>113</b> being preferably set to zero.

Each entry <b>113</b> is coupled to a succeeding entry <b>113</b> in the same column <b>111</b> but in the next row <b>112</b>. A bottom entry <b>113</b> in each column <b>111</b> in the last row <b>112</b> is coupled to a hashing element <b>120</b>, which determines a hash value in response thereto.

The hashing element <b>120</b> is coupled to an external memory <b>130</b>, which associates the hash value with a corresponding set of routing results. The information in the memory is responsive to the length field for the header information. The memory preferably includes a plurality of entries matching the header information and differing with regard to he length field. The length field preferably indicates a longest match for a destination address included in the header information. In a preferred embodiment, the external memory <b>130</b> includes a hash table with a set of entries associated with each hash value. For example, the external memory <b>130</b> can include an ordered list of entries associated with each hash value. Hash tables are known in the art of computer science.

In a preferred embodiment, hash values which are coupled to the external memory <b>130</b> are pipelined. When the external memory <b>130</b> is responding to a first hash value, the time delay is used to supply further hash values to the external memory <b>130</b> so as to take advantage of the intervening time between requesting data from the external memory <b>130</b> and receiving the requested data in response. After each entry <b>113</b> is coupled to the hashing element <b>120</b>, a next entry <b>113</b> is coupled to the hashing element <b>120</b> in succession, so that a plurality of k<b>2</b> entries <b>113</b> are pipelined and a similar plurality of k<b>2</b> hash values are supplied in parallel to the external memory <b>130</b> and are in various stages of response.

In a preferred embodiment, k<b>2</b> is at least five, so as to accommodate at least five lookup operations in parallel. It is expected that the external memory <b>130</b> takes at least about five times as long to respond to a lookup request as it takes for a single entry <b>113</b> to be processed. A leading parallel pointer <b>141</b> references a particular column <b>111</b> which has been earliest coupled to the external memory <b>130</b> for lookup. Similarly, a trailing parallel pointer <b>142</b> references a particular column <b>111</b> which has been latest coupled to the external memory <b>130</b> for lookup. The leading parallel pointer <b>141</b> and the trailing parallel pointer <b>142</b> are maintained k<b>2</b> columns <b>111</b> apart.

In a preferred embodiment, k<b>1</b> (the number of columns <b>111</b>) is about eleven, at least k<b>2</b> (five) columns <b>111</b> so as to be able to perform at least k<b>2</b> lookup operations can proceed in parallel, plus one additional column <b>111</b> so as to provide time for clearing one of the columns <b>111</b> after routing lookup is complete, plus about five additional columns <b>111</b> so as to provide room for inserting new sets of routing information.

An output of the external memory <b>130</b> is coupled to a comparison element <b>150</b>, which compares a result of the lookup of the hash value in the external memory <b>130</b> with the associated routing lookup information. The result of the lookup includes both a routing result and the associated routing lookup information. If the external memory <b>130</b> has more than one set of routing lookup information and routing results associated with a particular hash value, and responds with an incorrect association from the hash table, the comparison element <b>150</b> can note the difference and signal the external memory <b>130</b> to respond with a further association from the hash table.

Method of Operation

FIG. 2 shows a process flow diagram of a method of operating a system for multiple parallel packet routing lookup.

A method <b>200</b> of operating the system <b>100</b> is controlled by a memory controller <b>160</b>, which controls the hashing element <b>120</b>, the external memory <b>130</b>, the leading parallel pointer <b>141</b>, the trailing parallel pointer <b>142</b>, and the comparison element <b>150</b>, and other elements described herein, and includes flow points and process steps as described herein.

At a flow point <b>210</b>, one or more words <b>102</b> of routing lookup information have arrived at the input port <b>101</b>, and are ready for lookup.

The step <b>221</b> is performed in parallel with the steps <b>231</b> and following, which are performed in parallel with the steps <b>241</b> and following, so as to continuously perform lookup operations in parallel with the arrival of routing lookup information.

At a step <b>221</b>, a column <b>111</b> of the lookup holding table <b>110</b> is selected, and the words <b>102</b> of routing lookup information are entered into the selected first column <b>111</b> of the lookup holding table <b>110</b>. The column <b>111</b> is selected so as not to be between the leading parallel pointer <b>141</b> and the trailing parallel pointer <b>142</b>.

In a preferred embodiment, addresses having the most common bitmask length are inserted into the lookup holding table <b>110</b> so as to be addressed first, but so that more specific addresses are still considered when they are present.

It has been discovered that most internet addresses can be found by reference to 24-bit or 21-bit IP addresses, after which 16-bit, 12-bit, and finally 32-bit IP addresses are considered. When an IP address is to be entered into the external memory <b>130</b>, it is appended so as to meet one or more of these specific lengths.

When an IP address is to be searched in the external memory <b>130</b>, entries having each of these specific lengths, in this specific order, are searched. Thus, a 24-bit search value is entered at the bottom entry <b>113</b> of the column <b>111</b>, followed by a 21-bit search value, followed by a 16-bit search value, a 12-bit search value, and finally a 32-bit search value. When a relatively uncommon but more specific 32-bit IP address match is available, the entries for shorter search values are tagged with a “more specific value” tag to so indicate.

At a step <b>231</b>, a bottom entry <b>113</b> in each column <b>111</b> between the leading parallel pointer <b>141</b> and the trailing parallel pointer <b>142</b> is coupled to the hashing element <b>120</b>. The hashing element <b>120</b> generates a hash key for indexing into lookup tables in the external memory <b>130</b>. The memory controller <b>160</b> transmits routing lookup information from the bottom entry <b>113</b> for each column <b>111</b> in turn to the hashing element <b>120</b>, which in response generates a hash key and transmits that hash key to the external memory <b>130</b>.

At a step <b>232</b>, the hashing element <b>120</b> generates a hash key in response to the routing lookup information at the bottom of the selected second column <b>111</b>, and transmits that hash key to the external memory <b>130</b> for the lookup operation.

At a step <b>241</b>, the external memory <b>130</b> performs the lookup operation with the hash key.

In a preferred embodiment, the external memory <b>130</b> takes several (about five) clock cycles to respond with stored routing results, so the memory controller <b>160</b> pipeline processes a similar number (about five) hash values to the external memory <b>130</b> so as to pipeline process memory lookup by the external memory <b>130</b>. Pipeline processing is known in the art of computer processing.

At a step <b>242</b>, the external memory <b>130</b> couples a result of the lookup operation to the comparison element <b>150</b>.

In the step <b>242</b>, if the “more specific value” tag is set for the result of the lookup operation, the memory controller <b>160</b> records the result of the first routing lookup, flushes the search values up to the row <b>112</b> having the 32-bit search value, and retries the lookup as if the lookup operation had failed in the step <b>243</b>. Thereafter, if the lookup for the 32-bit search value succeeds, the memory controller <b>160</b> uses the result of the 32-bit routing lookup; otherwise, if the lookup for the 32-bit search value fails, the memory controller <b>160</b> uses the recorded result of the first routing lookup.

At a step <b>243</b>, the comparison element <b>150</b> compares the original routing lookup information with the result of the lookup operation. If the two match, the lookup operation is deemed successful, and the result of the lookup operation is deemed to include routing information for properly routing the packet. The routing information is forwarded to other devices so as to route the packet, and the method <b>200</b> continues with the step <b>244</b>.

If the two do not match, the lookup operation is deemed unsuccessful so far. In a preferred embodiment, the hash table in the external memory <b>130</b> includes a set of secondary entries for the particular hash key, preferably disposed as disclosed herein in a ordered list. An address for the next entry in the list is transmitted to the external memory <b>130</b> and the method <b>200</b> repeats the lookup operation until either a successful lookup operation occurs, or there are no further secondary entries, in which case the lookup operation fails.

At a step <b>244</b>, the lookup operation succeeded, and the remaining ones of the multiple entries <b>113</b> are flushed from the lookup holding table. The column <b>111</b> in which those multiple entries <b>113</b> becomes free for another set of multiple entries <b>113</b> from a packet. The method <b>200</b> continues at the flow point <b>210</b>.

At a step <b>245</b>, the lookup operation failed, and the remaining ones of the multiple entries <b>113</b> are each moved down one position, so that the second entry <b>113</b> becomes the bottom entry <b>113</b>. If there are no further ones of the multiple entries <b>113</b> for the particular packet, routing for the packet has failed and the packet is dropped. Whether or not the packet is dropped, the method <b>200</b> continues at the flow point <b>210</b>.

Alternative Embodiments

Although preferred embodiments are disclosed herein, many variations are possible which remain within the concept, scope, and spirit of the invention, and these variations would become clear to those skilled in the art after perusal of this application.